I'm working on some patches to move the 3c59x driver to the somewhat
new pci_iomap API. While straight-forward, this does hit ~250 places
in the code. So, it would be good to get some "just in case" testing
before I send it upstream.
Moving to the pci_iomap should also make it fairly simple to move to
using memory-mapped PCI I/O resources for those 3c59x devices that
support it. For the time being I have added a module option to the
3c59x driver called "use_mmio". Set that value to one (1) in order
to make use of this feature:
alias eth0 3c59x
options 3c59x use_mmio=1
What do I want from you? If you have 3c59x hardware, I'd like you
to try my test kernels at the locations below.
First, I'd like you to simply install the kernel and reboot. If the
3c59x driver does NOT work, then let me know.
If that works, then I'd like you to make the modprobe.conf
modifications described above and reboot. Please note whether or
not the device is still working. If it does work, please also note
if there is any noticeable difference in the performance of the
Finally, please send me the output of running "lspci -n", along with
the results of your testing. It would probably be best if you sent
them directly to me instead of flooding the list.
I appreciate your cooperation. Thanks in advance!
P.S. Test kernels for FC3, FC4, and Rawhide (aka FC5) available here:
John W. Linville
while working on a wrapper for gdl I ran into a problem with
Glib:ustring. The following "program" gives a compile error.
int main(int argc, char* argv)
The error is;
g++ -g -c `pkg-config gtkmm-2.4 --cflags` main.cpp
/usr/include/glibmm-2.4/glibmm/ustring.h:236: error: field initializer is not constant
make: *** [main.o] Error 1
The lines in ustring.h that the compiler complains about are;
235 #ifdef GLIBMM_HAVE_ALLOWS_STATIC_INLINE_NPOS
236 static const size_type npos = std::string::npos;
238 //The IRIX MipsPro compiler says "The indicated constant value is not known",
239 //so we need to initalize the static member data elsewhere.
240 static const size_type npos;
The following RPM are installed on my x86_64 (running in 64bit mode)
Anybody seen this problem before ? Can I just #undef
GLIBMM_HAVE_ALLOWS_STATIC_INLINE_NPOS somehow to "fix" it ?
On Thu, 2005-09-01 at 22:57 +0100, Chris Vine wrote:
> On Thursday 01 September 2005 21:45, Erwin Rol wrote:
> gcc-4.0 does allow static member variables to be initialised inline. So do
> gcc-3.2, 3.3 and 3.4, with x86 at any rate.
> Older pre-C++98 compilers may not. What does 'gcc --version' tell you? You
> may have more than one compiler installed and be calling the wrong one.
gcc (GCC) 4.0.1 20050831 (Red Hat 4.0.1-12)
and the system is;
Linux xpc.home.erwinrol.com 2.6.12-1.1519_FC5 #1 SMP Sat Aug 27 13:54:40 EDT 2005 x86_64 x86_64 x86_64 GNU/Linux
Its not the wrong compiler, i am sure about that.
And the problem is not that the member variables are initialized inline,
the problem is that std::string::npos is not constant, and so can not be
used to initialize static inline initialization.